基于预测的遗传算法  

Prediction in Genetic Algorithms

作　　者：陆锁军[1] 苗清影[1] 郭钊侠[1] 

机构地区：[1]东华大学信息科学与技术学院,上海201620

出　　处：《计算机仿真》2009年第3期183-186,共4页Computer Simulation

摘　　要：遗传算法被广泛应用于解决各类优化问题。常规的遗传算法易于陷入局部最优,其收敛速度也较慢。为了提高常规遗传算法的优化性能,将预测的概念引入遗传算法的循环过程,提出基于预测的遗传算法框架;并以人工神经网络算法作为预测算法,提出了一种基于神经网络预测的遗传算法。通过优化8个典型的函数优化问题,将该算法与常规遗传算法的性能进行了比较;结果显示该算法具有很强的全局优化能力,能有效地增强种群的多样性和进化速度,明显优于常规遗传算法。The genetic algorithm （GA） is adopted extensively to solve various optimization problems. The basic GA （BGA） is liable to local convergence and its convergence speed is very slow. To improve the convergence speed and the solution quality of the BGA, prediction is introduced into the cycled procedure of BGA and the framework of a prediction based GA is presented firstly. Taking the artifical neural network as the prediction algorithm, a neural network prediction based GA （NeuroPGA） is proposed. By optimizing eight typical function optimization problems, the performances of the NeuroPGA and a BGA are compared. The results show that the proposed algorithm can im- prove the global optimization capability and the diversity of the genetic population, also accelerate the evolutionary process, which is obviously superior to the BGA.

关 键 词：遗传算法 人工神经网络 预测 性能比较 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]